The Ethical Dilemma when (not) Setting up Cost-based Decision Rules in Semantic Segmentation

TL;DR

This paper explores the ethical implications of incorporating class-specific costs into semantic segmentation decision rules, highlighting how different cost functions impact safety-critical metrics.

Contribution

It introduces and compares egoistic and altruistic cost functions for semantic segmentation, emphasizing ethical considerations in decision rule design.

Findings

Cost functions significantly influence safety-related metrics.

Interpolating between MAP and cost-based rules alters false positive/negative rates.

Explicit cost specification raises ethical awareness in model deployment.

Abstract

Neural networks for semantic segmentation can be seen as statistical models that provide for each pixel of one image a probability distribution on predefined classes. The predicted class is then usually obtained by the maximum a-posteriori probability (MAP) which is known as Bayes rule in decision theory. From decision theory we also know that the Bayes rule is optimal regarding the simple symmetric cost function. Therefore, it weights each type of confusion between two different classes equally, e.g., given images of urban street scenes there is no distinction in the cost function if the network confuses a person with a street or a building with a tree. Intuitively, there might be confusions of classes that are more important to avoid than others. In this work, we want to raise awareness of the possibility of explicitly defining confusion costs and the associated ethical difficulties if it comes down to providing numbers. We define two cost functions from different extreme perspectives, an egoistic and an altruistic one, and show how safety relevant quantities like precision / recall and (segment-wise) false positive / negative rate change when interpolating between MAP, egoistic and altruistic decision rules.